CN115879357A - Self-adaptive bias proportion guidance method based on neural network - Google Patents

Self-adaptive bias proportion guidance method based on neural network Download PDF

Info

Publication number
CN115879357A
CN115879357A CN202111137643.7A CN202111137643A CN115879357A CN 115879357 A CN115879357 A CN 115879357A CN 202111137643 A CN202111137643 A CN 202111137643A CN 115879357 A CN115879357 A CN 115879357A
Authority
CN
China
Prior art keywords
neural network
initial
adaptive bias
angle
guidance method
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111137643.7A
Other languages
Chinese (zh)
Inventor
范世鹏
刘畅
王江
林德福
王因翰
刘经纬
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Institute of Technology BIT
Original Assignee
Beijing Institute of Technology BIT
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Institute of Technology BIT filed Critical Beijing Institute of Technology BIT
Priority to CN202111137643.7A priority Critical patent/CN115879357A/en
Publication of CN115879357A publication Critical patent/CN115879357A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation

Landscapes

  • Feedback Control In General (AREA)

Abstract

The invention discloses a self-adaptive bias proportion guidance method based on a neural network, which aims at a static fixed target and utilizes the neural network to obtain a constant item in bias proportion guidance, wherein the neural network is a BP (back propagation) neural network, the input of the neural network is the missile eye distance, the initial ballistic inclination angle, the initial missile eye sight angle and the expected terminal intersection angle when an aircraft is launched, and the output is the constant item. The self-adaptive bias proportion guidance method based on the neural network is high in guidance precision, can complete parameter solution of bias proportion guidance on line under different initial conditions and constraints, and is flexible to use and low in calculation cost.

Description

Self-adaptive bias proportion guidance method based on neural network
Technical Field
The invention relates to a self-adaptive bias proportion guidance method based on a neural network, and belongs to the technical field of aircraft control.
Background
The selection of the guidance law plays a key role in the accuracy of the terminal intersection angle.
At present, the bias proportional guidance method is widely applied, a constant bias term is added on the basis of the traditional proportional guidance law, and the method has the advantages of less required information amount, simple structure and the like, particularly, the residual time of flight does not need to be estimated, so the method has obvious advantages in engineering application.
However, the accuracy of the bias term directly affects the guidance accuracy, and the accuracy of the terminal intersection angle is affected by errors in the conventional formula due to the small angle assumption applied during derivation, the formula applied during total flight time estimation and the like.
In addition, the traditional formula derivation method needs professional technicians to temporarily calculate derivation bias items according to actual initial conditions, so that the use limitation is more, and the calculation derivation workload is larger.
For the above reasons, it is necessary to provide an offset proportional steering method with high accuracy that can be obtained quickly on-line.
Disclosure of Invention
In order to overcome the above problems, the present inventors have conducted intensive studies to design an adaptive bias proportion guidance method based on a neural network, and for a stationary fixed target, a constant term in bias proportion guidance is obtained by using the neural network.
Further, the neural network is a BP neural network.
In a preferred embodiment, the inputs of the neural network are the projectile distance r and the initial ballistic inclination angle theta of the aircraft at the time of launch 0 Initial bullet eye line of sight q 0 Angle of intersection with desired terminal f And the output of the BP neural network is a constant term b.
In a preferred embodiment, the steering law for the offset proportional steering is obtained according to the following equation:
Figure BDA0003282687870000021
wherein, theta is a trajectory inclination angle, q is a line-of-sight angle of the bullet, and N is a guidance coefficient.
In a preferred embodiment, the number of hidden layers of the BP neural network is 5.
In a preferred embodiment, the neural network needs to be trained before being used, and training samples are obtained by using ballistic simulation.
In a preferred embodiment, the trajectory simulation is a nonlinear percussion model simulation, which can be expressed as:
Figure BDA0003282687870000022
Figure BDA0003282687870000023
Figure BDA0003282687870000024
Figure BDA0003282687870000025
η＝θ-q
wherein eta is the included angle between the speed of the aircraft and the line of sight of the bullet eyes, theta is the inclination angle of the trajectory, q is the angle of the line of sight of the bullet eyes, r is the distance of the bullet eyes, v is the speed of the aircraft, a M Is an overload instruction for the aircraft.
In a preferred embodiment, when a training sample is established, the setting parameters of the trajectory simulation are modified for multiple times to obtain the terminal intersection angle under different setting parameters, wherein the setting parameters of the trajectory simulation comprise an initial projectile distance, a guidance coefficient, an initial trajectory inclination angle and an initial projectile line-of-sight angle.
In a preferred embodiment, the initial ballistic separation is 5000-10000 meters, the guidance factor is 2-4, and the initial ballistic dip angle is 10-20 °.
In a preferred embodiment, in the process of training the neural network, the Adam learning rate is used to update the parameters of the neural network.
The invention has the advantages that:
(1) The high-precision fitting approximation of the mapping is realized by adopting a BP neural network, and the guidance precision is high;
(2) The parameter solution of the bias proportion guidance can be completed on line under different initial conditions and constraints, and the use is flexible;
(3) The calculation amount required by the traditional formula in solving is reduced, and the calculation cost is reduced in engineering application.
Drawings
FIG. 1 is a flow chart of an adaptive bias proportion guidance method based on a neural network according to a preferred embodiment of the invention;
FIG. 2 shows a simulated ballistic trajectory of the lead law in example 1;
FIG. 3 shows the simulated ballistic dip of the guidance law in example 1;
fig. 4 shows simulated ballistic inclination angles in example 2 according to the present invention and comparative example 1.
Detailed Description
The invention is explained in more detail below with reference to the figures and examples. The features and advantages of the present invention will become more apparent from the description.
The word "exemplary" is used exclusively herein to mean "serving as an example, embodiment, or illustration. Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments. While the various aspects of the embodiments are presented in drawings, the drawings are not necessarily drawn to scale unless specifically indicated.
According to the self-adaptive bias proportion guidance method based on the neural network, provided by the invention, aiming at a static fixed target, bias proportion guidance is obtained by using the neural network, so that the accurate foot-landing constraint of an aircraft is realized.
The steering law of the bias ratio steering can be expressed as:
Figure BDA0003282687870000041
wherein, theta is a trajectory inclination angle, q is a line-of-sight angle of a bullet, N is a guidance coefficient, and b is an offset term.
In conventional offset proportional steering, the offset term is generally obtained by the following equation:
Figure BDA0003282687870000042
wherein, theta f Is the terminal intersection angle, θ 0 At an initial ballistic inclination, q 0 At an initial line of sight of the bullet eye, t 0 Is the aircraft initial launch time, t f The aircraft terminal rendezvous time.
For t f -t 0 The method is obtained by the following traditional method:
t＝t f -t 0
wherein t is the flight time taken by the aircraft to fly from the launch to the target location; due to t f -t 0 The small angle assumption and the estimation of the total time left in flight are applied, so that the error exists, and the finally obtained guidance law of the bias proportion guidance has low accuracy.
The inventors found that t is obtained by the conventional method f -t 0 The error in (b) can be compensated by adding an angular deviation term Δ η and a time deviation term Δ t:
Figure BDA0003282687870000043
wherein r is the projectile distance, v is the aircraft speed,
Figure BDA0003282687870000044
are all positive values, then | θ f I is a strictly monotone increasing function about b, and a one-to-one mapping relation exists between the strictly monotone functions, namely, theta f The one-to-one mapping relationship between | and b exists, and can be expressed as:
b＝f -1 (r,N,q 0 ,θ 0 ,θ f )
furthermore, the BP neural network in the neural network can predict the mapping relation more accurately, so that the bias proportion guidance is obtained by adopting the BP neural network in the invention.
According to the invention, the inputs of the neural network are the missile-mesh distance r and the initial trajectory inclination angle theta when the aircraft is launched 0 Initial bullet eye line of sight q 0 Angle of intersection with desired terminal f ；
Different from the constant term in the traditional method by the initial bullet sight angle q 0 Initial trajectory tilt angle theta 0 Desired terminal intersection angle θ f And the flight time t is obtained by resolving, in the invention, the flight time t with larger error is not used as the input of the neural network, and the bullet distance r which is easier to measure is used as the input of the neural network, thereby further improving the guidance precision.
Further, the output of the BP neural network is a constant term b, and a bias proportion guided guidance law is obtained according to the following formula:
Figure BDA0003282687870000051
wherein, theta is a trajectory inclination angle, q is a line-of-sight angle of the bullet, and N is a guidance coefficient.
In a preferred embodiment, the number of the hidden layers of the BP neural network is 5, and the inventor finds that, when there are 5 hidden layers, the guidance law obtained based on the constant term output by the neural network is more accurate, the calculation speed of the neural network is higher, and online solution can be realized.
Further preferably, the activation function of the BP neural network is a sigmod function:
Figure BDA0003282687870000052
further, the BP neural network needs to be trained before being used, and in the present invention, training samples are obtained by using ballistic simulation.
Specifically, the trajectory simulation is a nonlinear percussion model simulation, which may be expressed as:
Figure BDA0003282687870000061
Figure BDA0003282687870000062
Figure BDA0003282687870000063
Figure BDA0003282687870000064
η＝θ-q
wherein eta is the included angle between the speed of the aircraft and the line of sight of the bullet eyes, a M Is an overload instruction for the aircraft.
Further, the integral in the nonlinear percussion model is solved by using a Runge Kutta method, and preferably, the solving step size is 0.05.
The Runge-Kutta method (Runge-Kutta methods) is an important implicit or explicit iteration method for solving nonlinear ordinary differential equations, and is not described in detail in the present invention.
Further, when a training sample is established, the setting parameters of the trajectory simulation are modified for multiple times, and the terminal intersection angles under different setting parameters are obtained.
Further, the set parameters of the trajectory simulation comprise an initial bullet distance, a guidance coefficient, an initial trajectory inclination angle and an initial bullet sight angle.
In a preferred embodiment, the initial bullet mesh distance is 5000-10000 meters, the guidance factor is 2-4, and the initial ballistic inclination angle is 10-20 degrees in the set parameters.
In a preferred embodiment, a set of data is taken every 200m for the initial projectile distance, the guidance coefficients are 2, 3 and 4 respectively, and a set of data is taken at an initial ballistic inclination angle interval of 0.2 degrees as a set parameter of the ballistic simulation, so that a relatively comprehensive training sample is obtained.
In a preferred embodiment, the set parameters of the trajectory simulation in the training sample are adjusted according to the initial projectile distance, when the initial projectile distance is greater than 7500m, a group of data is taken every 100m, the guidance coefficients are 2, 3 and 4 respectively, and a group of data is taken at the initial trajectory inclination angle interval of 0.2 degrees as the set parameters of the trajectory simulation, so as to ensure that the trained neural network can obtain better accuracy;
when the initial missile mesh distance is less than 7500m, a group of data is taken at intervals of 200m, the guidance coefficients are respectively 2, 3 and 4, a group of data is taken at an initial ballistic inclination angle interval of 0.2 degrees and is used as a set parameter of ballistic simulation, and as the initial missile mesh distance is close, the sample size can be reduced properly on the premise of ensuring guidance precision, so that the calculation amount of a neural network is reduced, the resolving power of a missile-borne computer is reduced, namely, a high-configuration missile-borne computer is not needed, and the cost is reduced.
According to the invention, in the course of training the neural network, 60% -80% of the samples are selected from the training samples as the training set, for example, 70% of the samples are selected as the training set, 10% -20% of the samples are selected as the testing set, for example, 15% of the samples are selected as the testing set, and 10% -20% of the samples are selected as the verification set, for example, 15% of the samples are selected as the verification set.
Further preferably, in the process of training the neural network, parameters of the neural network are updated by using Adam learning rate, and the updating process may be represented as:
Figure BDA0003282687870000071
Figure BDA0003282687870000072
wherein,
Figure BDA0003282687870000073
is a pre-update parameter;
Figure BDA0003282687870000074
Is an updated parameter; kappa is the learning rate; ε is a smoothing term, preventing division by zero; m' t Is an estimate of the first moment, v' t Is a second moment estimation; m is t Is the first moment of the gradient, v t Is a gradient second moment, beta 1 、β 2 Is a constant exponential decay rate.
According to the invention, the initial trajectory inclination angle at the time of launching of the aircraft is preferably 10-20 °, and the inventor finds that the initial trajectory inclination angle has a great influence on the accuracy of the terminal intersection angle (i.e. the error between the actual terminal intersection angle and the expected terminal intersection angle), and when the initial trajectory inclination angle is too small, the accuracy of the terminal intersection angle is obviously reduced, and when the initial trajectory inclination angle is too large, the accuracy of the terminal intersection angle is also affected, and more preferably, the initial trajectory inclination angle is 15-20 °, and within the range, the accuracy of the terminal intersection angle is the best.
Examples
Example 1
Obtaining self-adaptive bias proportion guidance based on a neural network, wherein the guidance law of the bias proportion guidance is as follows:
Figure BDA0003282687870000081
and (3) acquiring a constant term b in the bias proportion guidance by adopting a BP (back propagation) neural network, wherein the number of hidden layers of the neural network is 5, and the activation function is a sigmod function.
The training sample is obtained through trajectory simulation, the trajectory simulation is nonlinear striking model simulation, and the nonlinear striking model is as follows:
Figure BDA0003282687870000082
Figure BDA0003282687870000083
Figure BDA0003282687870000084
Figure BDA0003282687870000085
η＝θ-q
the integral in the nonlinear hit model is solved by adopting a Runge Kutta method, the solving step length is 0.05, the set parameters of trajectory simulation comprise an initial bullet-target distance, a guidance coefficient, an initial trajectory inclination angle and an initial bullet-target sight angle, wherein the initial bullet-target distance is 5000-10000 m, the guidance coefficient is 2-4, the initial trajectory inclination angle is 10-20 degrees, a set of data is taken at the initial bullet-target distance every 200m, the guidance coefficients are 2, 3 and 4 respectively, and a set of data is taken at the initial trajectory inclination angle interval of 0.2 degrees.
From the training samples, 70% of the samples were selected as the training set, 15% of the samples were selected as the test set, and 15% of the samples were selected as the validation set.
In the process of training the neural network, parameters of the neural network are updated by adopting the Adam learning rate, and the updating process is represented as:
Figure BDA0003282687870000091
Figure BDA0003282687870000092
simulating by adopting the trained BP neural network, wherein the expected terminal intersection angle theta in simulation parameters f Set to-30 °, -60 °, -90 °, respectively, and the remaining simulation parameters are as shown in table one:
watch 1
Figure BDA0003282687870000093
According to the parameters, the input parameters of the neural network are as follows: projectile distance r =10000, initial trajectory inclination angle θ 0 =10 DEG initial bullet eye sight angle q 0 =0 °, desired terminal intersection angle θ f Respectively at-30 deg., 60 deg., and 90 deg..
Obtaining an offset proportion guidance law according to a BP neural network, wherein a simulated ballistic trajectory of the guidance law is shown in FIG. 2, a ballistic inclination angle is shown in FIG. 3, and it can be seen from the diagram that in all three cases, an aircraft can reach a target position, and actual terminal intersection angles are respectively: 29.9908 °, -60.0016 °, -89.9976 °, with a small angular difference from the desired terminal intersection.
Example 2
The simulation was performed by the same method as in example 1 except that, in the simulation parameters, the terminal intersection angle θ was expected f Set to-40 deg. -55 deg., respectively, and the rest simulation parameters are shown in table two:
watch two
Figure BDA0003282687870000094
The trajectory of the trajectory inclination obtained by the simulation is shown in fig. 4, with the actual terminal intersection angles of the aircraft being-40.21 and-54.97 deg., respectively.
Comparative example 1
The same simulation experiment as that of the embodiment 2 is carried out, except that a traditional formula is adopted to solve and obtain a guidance law of bias proportion guidance, the simulated trajectory inclination angle of the guidance law is shown in figure 4, and the actual terminal intersection angles of the aircraft are-41.95-61.14 degrees respectively.
Comparing the simulated ballistic inclination angles of the example 2 and the comparative example 1, it can be seen that the guidance law of the bias proportion guidance obtained in the example 2 is higher in accuracy, the error between the actual intersection angle and the expected intersection angle is reduced by more than 9 times, especially when the intersection angle of the expected terminal is large, the error between the actual intersection angle and the expected intersection angle is reduced by nearly 38 times, the error between the actual intersection angle and the expected intersection angle is as low as 0.05%, and the guidance accuracy improvement effect is obvious.
In the description of the present invention, it should be noted that the terms "upper", "lower", "inner", "outer", "front", "rear", and the like indicate orientations or positional relationships based on operational states of the present invention, and are only used for convenience of description and simplification of description, but do not indicate or imply that the referred device or element must have a specific orientation, be constructed in a specific orientation, and be operated, and thus should not be construed as limiting the present invention. Furthermore, the terms "first," "second," "third," and "fourth" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.
In the description of the present invention, it should be noted that, unless otherwise specifically stated or limited, the terms "mounted," "connected," and "connected" are to be construed broadly, e.g., as meaning either a fixed connection, a removable connection, or an integral connection; can be mechanically or electrically connected; the connection may be direct or indirect via an intermediate medium, and may be a communication between the two elements. The specific meanings of the above terms in the present invention can be understood in specific cases to those skilled in the art.
The present invention has been described above in connection with preferred embodiments, but these embodiments are merely exemplary and merely illustrative. On the basis of the above, the invention can be subjected to various substitutions and modifications, and the substitutions and the modifications are all within the protection scope of the invention.

Claims (10)

1. A self-adaptive bias proportion guidance method based on a neural network aims at a static fixed target and utilizes the neural network to obtain a constant term in bias proportion guidance.
2. The neural network-based adaptive bias proportion guidance method according to claim 1,
the neural network is a BP neural network.
3. The neural network-based adaptive bias proportion guidance method according to claim 1,
the inputs of the neural network are the projectile distance r and the initial trajectory inclination angle theta when the aircraft is launched 0 Initial line of sight q of the bullet 0 Angle of intersection with desired terminal f And the output of the BP neural network is a constant term b.
4. The neural network-based adaptive bias proportion guidance method according to claim 1,
the steering law for the offset proportional steering is obtained according to the following formula:
Figure FDA0003282687860000011
wherein, theta is a trajectory inclination angle, q is a line-of-sight angle of the bullet, and N is a guidance coefficient.
5. The neural network-based adaptive bias proportion guidance method according to claim 2,
the number of the hidden layers of the BP neural network is 5.
6. The neural network-based adaptive bias proportion guidance method according to claim 1,
the neural network needs to be trained before being used, and training samples are obtained by adopting trajectory simulation.
7. The neural network-based adaptive bias proportion guidance method according to claim 6,
the trajectory simulation is a nonlinear percussion model simulation, which can be expressed as:
Figure FDA0003282687860000012
Figure FDA0003282687860000021
Figure FDA0003282687860000022
Figure FDA0003282687860000023
η＝θ-q
wherein eta is the included angle between the speed of the aircraft and the line of sight of the bullet eyes, theta is the inclination angle of the trajectory, q is the angle of the line of sight of the bullet eyes, r is the distance of the bullet eyes, v is the speed of the aircraft, a M Is an overload instruction for the aircraft.
8. The neural network-based adaptive bias proportion guidance method according to claim 6,
when a training sample is established, modifying the set parameters of the trajectory simulation for multiple times to obtain terminal intersection angles under different set parameters, wherein the set parameters of the trajectory simulation comprise an initial bullet distance, a guidance coefficient, an initial trajectory inclination angle and an initial bullet sight angle.
9. The neural network-based adaptive bias proportion guidance method according to claim 6,
the initial missile distance is 5000-10000 m, the guidance coefficient is 2-4, and the initial ballistic inclination angle is 10-20 degrees.
10. The neural network-based adaptive bias proportion guidance method according to claim 1,
in the process of training the neural network, parameters of the neural network are updated by adopting the Adam learning rate.
CN202111137643.7A 2021-09-27 2021-09-27 Self-adaptive bias proportion guidance method based on neural network Pending CN115879357A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111137643.7A CN115879357A (en) 2021-09-27 2021-09-27 Self-adaptive bias proportion guidance method based on neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111137643.7A CN115879357A (en) 2021-09-27 2021-09-27 Self-adaptive bias proportion guidance method based on neural network

Publications (1)

Publication Number Publication Date
CN115879357A true CN115879357A (en) 2023-03-31

Family

ID=85763051

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111137643.7A Pending CN115879357A (en) 2021-09-27 2021-09-27 Self-adaptive bias proportion guidance method based on neural network

Country Status (1)

Country Link
CN (1) CN115879357A (en)

Similar Documents

Publication Publication Date Title
US10502527B2 (en) Real-time ballistic solutions for calculating an aiming adjustment and for indicating a subsonic threshold
DK156333B (en) PROCEDURE FOR TRACKING THE POSITION OF A MOVABLE TARGET
US10415933B1 (en) Real-time ballistic solutions for moving-target aiming calculations
CN115248038B (en) SINS/BDS combined navigation engineering algorithm under emission system
CN111351401A (en) Anti-sideslip guidance method applied to strapdown seeker guidance aircraft
CN105628051A (en) Embedded atmosphere measuring equipment performance evaluation method
CN115879357A (en) Self-adaptive bias proportion guidance method based on neural network
NO322766B1 (en) Neural network for path command controller
US8770481B2 (en) Ballistic nomograph for small arms fire
CN109084772B (en) Unscented Kalman based sight line conversion rate extraction method and system
CN116466744A (en) Guidance method with view angle and falling angle constraint suitable for variable speed aircraft
CN111043914B (en) Method and system for acquiring bullet sight line angular rate applied to strapdown seeker
CN110082749A (en) Shell outer trajectory state of flight estimation method
CN111649734B (en) Particle swarm algorithm-based strapdown seeker target positioning method
CN111290436B (en) Aircraft wireless instruction correction method and system
CN115046433A (en) Aircraft time collaborative guidance method based on deep reinforcement learning
Xin et al. Terminal guidance simulation and flight test for small UCAV
CN114771877B (en) Optimal interception guidance method considering navigation error
CN117889868B (en) Missile position accurate estimation method integrating infrared seeker information
RU2782035C2 (en) Simulation model of system for control of air target based on unmanned aerial vehicle of target complex
CN112306073B (en) Multi-lead-angle guiding method based on fuzzy confidence superposition
CN114967749B (en) Maneuvering trajectory design method for low-cost altimeter
CN117195711A (en) Bullet-eye distance and residual flight time analysis method based on data driving
CN113064443B (en) Gain online adjustment method and damping loop control method using same
Hao et al. Aircraft guidance instrumentation error estimation based on neural network method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination